Fundamentals of Database Design

Fundamentals of Database Design Zornitsa Zaharieva CERN Data Management Section - Controls Group Accelerators and Beams Department /AB-CO-DM/ 23-FEB-2005

Contents : Introduction to Databases : Main Database Concepts : Conceptual Design : Entity-Relationship Model : Logical Design : Relational Model : Introduction to SQL : Implementing the Relational Model through DDL : Best Practices in Database Design 2/50

3/50

Databases - Evolution Data stored in file systems problems with : redundancy : maintenance : security : efficient access to the data Database Management Systems Software tools that enable the management (definition, creation, maintenance and use) of large amounts of interrelated data stored in a computer accessible media. 1 st generation of Database Management Systems : based on hierarchical and network models 2 nd generation of DBMS : 1969 Dr. Codd proposed the relational model 4/50

Capabilities of a Database Management System Manage persistent data Access large amounts of data efficiently Support for at least one data model Support for certain high-level language that allow the user to define the structure of the data, access data, and manipulate data Transaction management the capability to provide correct, concurrent access to the database by many users at once Access control the ability to limit access to data by unauthorized users, and the ability to check the validity of data Resiliency the ability to recover from system failures without losing data 5/50

Data Model A mathematical abstraction (formalism) through which the user can view the data Has two parts 1. A notation for describing data 2. A set of operations used to manipulate that data Examples of data models : relational model : network model : hierarchical model : object model 6/50

Design Phases Difficulties in designing the DB s effectively brought design methodologies based on data models Database development process Business Information Requirements Conceptual Design Produces the initial model of the real world in a conceptual model Logical Design Consists of transforming the conceptual schema into the data model supported by the DBMS Physical Design Aims at improving the performance of the final system Conceptual Data Modeling Logical Database Design Physical Database Design 7/50 Operational Database

Conceptual Design The process of constructing a model of the information used in an enterprise Is a conceptual representation of the data structures Is independent of all physical considerations Should be simple enough to communicate with the end user Should be detailed enough to create the physical structure Business information requirements Conceptual Design Conceptual model (Entity-Relationship Model) 8/50

Information Requirements CERN Controls Example There is a need to keep an index of all the controls entities and their parameters coming from different controls systems. Each controls entity has a name, description and location. For every entity there might be several parameters that are characterized by their name, description, unit, quantity code, data type and system they are sent from. This database will be accessed and exchange data with some of the existing databases related to the accelerators controls. It will ensure that every parameter name is unique among all existing controls systems. Naming db 9/50 Zornitsa Zaharieva Zornitsa CERN Zaharieva /AB-CO-DM/ CERN /AB-CO-DM/

Information Requirements CERN Controls Example Samples of the data that has to be stored: controls_entity name: VPIA.10020 description: Vacuum Pump Sputter Ion type A in location 10020 entity_code: VPIA expert_name: VPIA_10020 accelerator: SPS location_name: 10020 location_class: SPS_RING_POS location_class_description: SPS Ring position entity_parameter name: VPIA.10020:PRESSURE description: Pressure of Vacuum Pump Sputter Ion type A in location 10020 expert_name: VPIA.10020.PR unit_id: mb unit_description: millibar data_type: NUMERIC quantity_code: PRESSURE system_name: SPS_VACUUM system_description: SPS Vacuum 10/50

Entity-Relationship Model The Entity-Relationship model (ER) is the most common conceptual model for database design nowadays No attention to efficiency or physical database design Describes data as entities, attributes, and relationships It is assumed that the Entity-Relationship diagram will be turned into one of the other available models during the logical design Entity-relationship model Hierarchical model Network model Relational model 11/50

Entity A thing of significance about which the business needs to store information trivial example: employee, department CERN controls example: controls_entity, location, entity_parameter, system, quantity_code, data_type Entity instance an individual occurrence of a given entity a thing that exists and is distinguishable J. Ullman Local Database /cerndb1/ Remote Database /edmsdb/ trivial example: a single employee CERN controls example: a given system (e.g. SPS Vacuum) Note: Be careful when establishing the boundaries for the entity, e.g. entity employee all employees in the company or all employees in a given department depends on the requirements 12/50

Attributes Attributes are properties which describe the entity attributes of system - name, description Attributes associate with each instance of an entity a value from a domain of values for that attribute set of integers, real numbers, character strings Attributes can be : optional : mandatory SYSTEM id description A Key - an attribute or a set of attributes, whose values uniquely identify each instance of a given entity 13/50

ER Modeling Conventions 14/50 If you use Oracle Designer the following convention is used: ENTITY Soft box Singular name Unique Uppercase ENTITY_PARAMETER # id * description o expert_name * unit_id * unit_description attribute Singular name Unique within the entity Lowercase Mandatory (*) Optional (o) Unique identifier (#) Note: There are different conventions for representing the ER model!

15/50 Relationships Associations between entities examples: employees are assigned to departments entity_parameters are generated by systems Degree - number of entities associated with a relationship (most common case - binary) Cardinality - indicates the maximum possible number of entity occurrences Existence - indicates the minimum number of entity occurrences set of integers, real numbers, character strings : mandatory : optional SYSTEM # id * description produces is generated by ENTITY_PARAMETER # id * description o expert_name

Relationship Cardinality One-to-One (1:1) one manager is a head of one department Note: Usually this is an assumption about the real world that the database designer could choose to make or not to. One-to-Many (1:N) one system could generate many parameters one parameter is generated by only one system Many-to-Many (N:M) many employees are assigned to one project one employee is assigned to many projects 16/50

ER Modeling Conventions If you use Oracle Designer the following convention is used: Relationship Name descriptive phrase Line connecting to entities Mandatory - solid line Optional - dashed line One - single line Many - crow s foot Note: There are different conventions for representing the ER model! 17/50

CERN Controls Example Entity-Relationship Diagram 18/50

Logical Design Business Information Requirements Conceptual Data Modeling Logical Database Design Physical Database Design 19/50 Translate the conceptual representation into the logical data model supported by the DBMS Conceptual model (Entity-Relationship Model) Operational Database Logical Design Normalized Relational Model

Relational Model The most popular model for database implementation nowadays Supports powerful, yet simple and declarative languages with which operations on data are expressed Value-oriented model Represents data in the form of relations Data structures relational tables Data integrity tables have to satisfy integrity constraints 20/50 Relational database a collection of relations or two-dimensional tables

Relational Table Composed by named columns and unnamed rows The rows represent occurrences of the entity Every table has a unique name Columns within a table have unique names Order of columns is irrelevant Every row is unique Order of rows is irrelevant Every field value is atomic (contains a single value) 21/50

Primary Key (PK) A column or a set of columns that uniquely identify each row in a table Composite (compound) key Role to enforce integrity : every table must have a primary key For every row the PK : must have a non-null value : the value must be unique : the value must not change or become null during the table lifetime Columns with these characteristics are candidate keys 22/50

Foreign Key (FK) Column(s) in a table that serves as a PK of another table Enforces referential integrity by completing an association between two tables 23/50

Data Integrity Refers to the accuracy and consistency of the data by applying integrity constraints rules Attributes associate with each instance of an entity a value from a domain of values for that attribute Constraint type Explanation Entity Integrity No part of a PK can be NULL ---------------------------------------------------------------------------------------------------------------- Referential Integrity A FK must match an existing PK value or else be NULL ---------------------------------------------------------------------------------------------------------------- Column Integrity A column must contain only values consistent with the defined data format of the column ---------------------------------------------------------------------------------------------------------------- User-defined Integrity The data stored in the database must comply with the business rules 24/50

From Entity-Relationship Model to Relational Model Entity-Relationship model Entity Attribute Key Relationship Relational model Relational table Column (attribute) Primary Key (candidate keys) Foreign Key SYSTEM SYSTEMS # id * description PK SYS_ID SYS_DESCRIPTION 25/50

Relationships Transformations Binary 1:1 relationships Solution : introduce a foreign key in the table on the optional side Binary 1:N relationship Solution : introduce a foreign key in the table on the many side M:N relationships Solution : create a new table; : introduce as a composite Primary Key of the new table, the set of PKs of the original two tables 26/50

CERN Controls Example Relational Model before normalization 27/50

Normalization A series of steps followed to obtain a database design that allows for consistent storage and avoiding duplication of data A process of decomposing relationships with anomalies The normalization process passes through fulfilling different Normal Forms A table is said to be in a certain normal form if it satisfies certain constraints Originally Dr. Codd defined 3 Normal Forms, later on several more were added 28/50

Normalization Relational db model Normalization process 1 st Normal Form 2 nd Normal Form For most practical purposes databases are considered normalized if they adhere to 3 rd Normal Form 3 rd Normal Form Boyce/Codd Normal Form 4 th Normal Form 5 th Normal Form Normalized relational db model 29/50

1 st Normal Form 1 st Normal Form - All table attributes values must be atomic : multi-values are not allowed By definition a relational table is in 1 st Normal Form Definition: functional dependency (A -> B) If attribute B is functionally dependent on attribute A, then for every instance of A you can determine the value of B 30/50

2 nd Normal Form 2 nd Normal Form - Every non-key attribute is fully functionally dependent on the PK : no partial dependencies : every attribute must be dependent on the entire PK LOCATIONS(lc_class_id, lc_name, lc_class_description) Solution: : for each attribute in the PK that is involved in a partial dependency, create a new table : all attributes that are partially dependent on that attribute should be moved to the new table LOCATIONS (loc_class_id, loc_name) LOCATION_CLASSES (lc_class_id, lc_class_description) 31/50

3 nd Normal Form No transitive dependencies for non-key attributes Definition: Transitive dependence When a non-key attribute depends on another non-key attribute. ENTITY_PARAMETERS(ep_id,,unit_id, unit_description) 32/50 Solution: : for each non-key attribute A that depends upon another non-key attribute B create a new table : create PK of the new table as attribute B : create a FK in the original table referencing the PK of the new table ENTITY_PARAMETERS(ep_id,,unit_id) UNITS(unit_id, unit_descrption)

Denormalization Queries against a fully normalized database often perform poorly Explanation: Current RDBMSs implement the relational model poorly. A true relational DBMS would allow for a fully normalized database at the logical level, whilst providing physical storage of data that is tuned for high performance. Two approaches are used Approach 1: Keep the logical design normalized, but allow the DBMS to store additional redundant information on disk to optimize query response (indexed views, materialized views, etc.). In this case it is the DBMS software's responsibility to ensure that any redundant copies are kept consistent. 33/50

Denormalization Approach 2: Use denormalization to improve performance, at the cost of reduced consistency Denormalization is the process of attempting to optimize the performance of a database by adding redundant data This may achieve (may not!) an improvement in query response, but at a cost There should be a new set of constraints added that specify how the redundant copies of information must be kept synchronized 34/50

Denormalization Denormalization can be hazardous : increase in logical complexity of the database design : complexity of the additional constraints It is the database designer's responsibility to ensure that the denormalized database does not become inconsistent 35/50

CERN Controls Example Relational Model after normalization 36/50

Structured Query Language Most commonly implemented relational query language SQL originally developed by IBM Used to create, manipulate and maintain a relational database Official ANSI standard 37/50

Structured Query Language Data Definition Language (DDL) : define the database schema : CREATE, DROP, ALTER table Data Manipulation Language (DML) : manipulate the data in the tables : SELECT, INSERT, UPDATE, DELETE Data Control Language (DCL) : control user access to the database schema : GRANT, REVOKE user privileges 38/50

Database schema implementation Definition: Database schema a collection of logical structures of data The implementation of the database schema is realized through the DDL part of SQL Although there is a standard for SQL, there might be some features when writing the SQL scripts that are vendor specific 39/50 Some commercially available RDBMS : Oracle : DB2 IBM : Microsoft SQL Server : Microsoft Access : mysql

Create Table Describe the layout of the table : table name : column names : datatype for each column : integrity constraints - column constraints, default values, not null -PK, FK CREATE TABLE systems ( sys_id,sys_description ); VARCHAR2(20) VARCHAR2(100) 40/50

Datatypes Each attribute of a relation (column in a table) in a RDBMS has a datatype that defines the domain of values this attribute can have The datatype for each column has to be specified when creating a table ANSI standard Oracle specific implementation 41/50

Oracle Datatypes CHAR (size) fixed-length char array VARCHAR2(size) variable-length char string NUMBER (precision, scale) any numeric DATE date and time with seconds precision TIMESTAMP data and time with nano-seconds precision CLOB char large object BLOB binary large object BINARY_FLOAT 32 bit floating point BINARY_DOUBLE 64 bit floating point + some others 42/50

Constraints Primary Key ALTER TABLE systems ADD( CONSTRAINT SYSTEM_PK PRIMARY KEY (sys_id)); Foreign Key ALTER TABLE entity_parameters ADD (CONSTRAINT EP_SYS_FK FOREIGN KEY (system_id) REFERENCES systems(sys_id)) Unique Key ALTER TABLE entity_parameters ADD (CONSTRAINT EP_UNQ UNIQUE (ep_name)); 43/50

Data Definition Language Statements Statements in the DDL : used for tables and other objects (views, sequences, etc.) CREATE ALTER DROP RENAME TRUNCATE CREATE SEQUENCE EP_SEQ NOMAXVALUE NOMINVALUE NOCYCLE NOCACHE 44/50

Best Practices in Database Design Black box syndrome : understand the features of the database and use them Relational database or a data dump : let the database enforce integrity : using the power of the relational database manage integrity in multi-user environment : using PK and FK : not only one application will access the database : implementing constraints in the database, not in the client or in the middle tier, is faster : using the right datatypes 45/50 Database independence

Best Practices in Database Design Not using generic database models : tables - objects, attributes, object_attributes, links : performance problem! Designing to perform Creating a development (test) environment Testing with real data and under real conditions 46/50

Best Practices in Database Design 47/50

Development Tools Oracle provided tools : Oracle Designer : SQL* Plus : JDeveloper Benthic Software - http://www.benthicsoftware.com/ : Golden : PL/Edit : GoldView : at CERN - G:\Applications\Benthic\Benthic_license_CERN.html Microsoft Visio CAST - http://www.castsoftware.com/ : SQL Code-Builder 48/50

References [1] Ensor, D., Stevenson, I., Oracle Design, O Reilly, 1997 [2] Kyte, T., Effective Oracle by Design [3] Loney, K., Koch, G., Oracle 9i The Complete Reference, McGraw-Hill, 2002 [4] Oracle course guide, Data Modeling and Relational Database Design, Oracle, 1996 [5] Rothwell, D., Databases: An Introduction, McGraw-Hill, 1993 [6] Ullman, J., Principles of Databases and Knowledge-Base Systems volumn 1, Computer Science Press, 1988 [7] Oracle on-line documentation http://oracle-documentation.web.cern.ch/oracle-documentation/ 49/50

End; Thank you for your attention! Zornitsa.Zaharieva@cern.ch 50/50